Awesome Chris! I'm hoping to get more of my coworkers on the Julia train but one of the first questions I always get from the grey-beards is "what are these solvers? how can I trust them and where is my ODE45." This talk is exactly what was needed!
When I started looking at Julia, one of the first things that impressed me was the richness and care put into its ODE solvers. While I have yet to use them in anger, this talk gives some context to my first impressions. Thank you.
I'm just at ~9:30 and the thought is about how Apollo got to the moon (3d navigation) with Quaternions where errors are pushed into an unimportant variable (diameter of the 'sphere', 4th dim) when what they needed was vector direction/orientation of the sphere. Diameter can be re-normalised later. IIRC they could handle 10 degree steps and still maintain amazing integrated direction control.
So good! Would be nice if you would spend like 30s to explain what e.g. stiff means (for noobs like me) but otherwise it was a really nice presentation. I was in the rk4 and chill camp but now I‘m higher order curious
@@aaronkaw4857 Interesting! That makes it sound like a quantitative property (blurred lines on what constitutes a "small stepsize"). From the name it always appeared to be a qualitative differrence between systems 🤔
Awesome Chris! I'm hoping to get more of my coworkers on the Julia train but one of the first questions I always get from the grey-beards is "what are these solvers? how can I trust them and where is my ODE45." This talk is exactly what was needed!
When I started looking at Julia, one of the first things that impressed me was the richness and care put into its ODE solvers. While I have yet to use them in anger, this talk gives some context to my first impressions. Thank you.
I'm just at ~9:30 and the thought is about how Apollo got to the moon (3d navigation) with Quaternions where errors are pushed into an unimportant variable (diameter of the 'sphere', 4th dim) when what they needed was vector direction/orientation of the sphere. Diameter can be re-normalised later.
IIRC they could handle 10 degree steps and still maintain amazing integrated direction control.
Great talk, thank you, Chris!
Good to watch, but do yourself a favor and lower the playback speed to 75% because he *talks so, so fast*.
So good! Would be nice if you would spend like 30s to explain what e.g. stiff means (for noobs like me) but otherwise it was a really nice presentation. I was in the rk4 and chill camp but now I‘m higher order curious
Stiffness is when your ode solver needs a really small stepsize, otherwise it's unstable.
@@aaronkaw4857 Interesting! That makes it sound like a quantitative property (blurred lines on what constitutes a "small stepsize"). From the name it always appeared to be a qualitative differrence between systems 🤔
I hope Tsitouris sees this presentation, or at least knows his paper is being applied.
he has very high blood pressure 😅